Video Ad Shot List from VSL

I want to build an AI agent that acts as a Video Ad Shot Planner and Creative Director Assistant. This agent will help me generate a detailed shot list and creative prompts for producing video ads. The user (me) will provide three inputs: 1. A Video Sales Letter (VSL) script (as plain text). 2. The Product Name. 3. A Product Description (including target audience, brand voice, and key selling points). The AI agent should perform the following sequence of tasks: 1. **Analyze VSL and Segment:** * Ingest the VSL script (plain text). * Using a powerful LLM (like Skill #223), intelligently break down the VSL script into key scenes, distinct messaging segments, or even impactful phrases/sentences. The LLM should use its creative understanding to identify these key moments suitable for distinct visuals. * For each identified segment, **clearly extract and reference the exact VSL script text** that defines that segment. This reference is crucial for the user to understand where each visual will be placed. 2. **Generate Visual Prompts per Segment:** * For each VSL segment identified: * Leveraging the VSL segment text, the product name, and the product description, use an LLM (like Skill #223) to generate **2-3 highly detailed and distinct visual prompt ideas**. * These prompts must be suitable for direct use in third-party AI image and video generation tools. They should creatively describe the imagined scene, camera angle, mood, key elements, lighting, and how the product might be featured or symbolized. The LLM should aim for creative and engaging visual concepts. * Clearly associate each set of visual prompts with its corresponding VSL script segment text reference. 3. **Suggest Music and Sound Design:** * Based on the overall VSL script, product name, and product description, use an LLM (like Skill #223 or #218 for keyword brainstorming) to: * Suggest 2-3 suitable music genres, moods, or types of royalty-free tracks (e.g., "energetic pop," "inspiring orchestral," "quirky ukulele") that would enhance the ad's emotional impact. * Optionally, suggest where simple sound effects might enhance the video, linking these to specific thematic moments if possible. 4. **Recommend Transitions:** * Use an LLM (like Skill #223) to suggest appropriate video transition types (e.g., "hard cut," "fade," "J-cut," "L-cut," "wipe") between the major visual segments/scenes to ensure a smooth, professional, and engaging final video ad. The suggestions should consider the pacing and mood of the VSL. 5. **Compile and Output:** * The final output should be a **detailed text document**. * The document should be structured clearly: For each VSL segment: * **VSL Script Reference:** Display the exact VSL script text for that segment. * **Visual Prompt Idea 1:** [Detailed prompt] * **Visual Prompt Idea 2:** [Detailed prompt] * **(Optional) Visual Prompt Idea 3:** [Detailed prompt] * **Suggested Transition:** [Transition type leading from this segment's visual to the next, if applicable]. * The output should also include a separate, clearly marked section detailing the overall music and sound design suggestions. The goal is to provide the user with a comprehensive creative brief and a series of actionable prompts they can use with other AI tools to generate all the visual assets needed for their video ad. The output must make it absolutely clear which VSL script portion each visual idea corresponds to, enabling easy assembly of the final video ad.

First, pre-entered info...

final-output2

The final deliverable is a single text-based document (for example, "VideoAdCreativeBrief.txt") that outlines how to produce the video ad. It does not include any actual media files.

subagents-extracted2

OVERVIEW OF SUBAGENTS • Subagent 1: VSL Script Segmenter • Subagent 2: Visual Prompt Generator • Subagent 3: Transitions & Music Advisor • Subagent 4: Document Compiler SUBAGENT 1: VSL Script Segmenter Name: VSL Script Segmenter Skills Used: • #223 (Powerful LLM Prompt-to-Text Response) Description: Takes the full VSL script as input and breaks it down into up to around 6 distinct segments. Each segment is labeled ("Segment 1," "Segment 2," etc.) and includes the exact script text. Input Needed: • Full VSL script (plain text). • Instruction to identify main messaging or emotional beats. Output: • Labeled segments, each with the exact script text (single structured output). SUBAGENT REQUIRED INPUT: {vsl-script} SUBAGENT FINAL OUTPUT: [segmented-vsl-script] SUBAGENT 2: Visual Prompt Generator Name: Visual Prompt Generator Skills Used: • #223 (Powerful LLM Prompt-to-Text Response) Description: Reads each VSL segment, the product name, and product description, then generates 2–3 creative visual prompt ideas per segment. These prompts describe scene, camera angle, lighting, mood, and how the product might appear, in concise form. Input Needed: • Output from Subagent 1 (labeled segments). • Product name and product description. Output: • A single structured output array of 2–3 visual prompts per labeled segment. SUBAGENT REQUIRED INPUT: [segmented-vsl-script], {product-name}, {product-description} SUBAGENT FINAL OUTPUT: [visual-prompts] SUBAGENT 3: Transitions & Music Advisor Name: Transitions & Music Advisor Skills Used: 1) #218 (Brainstorm Related Keywords) Description: Briefly reviews the product description to brainstorm 2–3 relevant music/sound-related keywords. Input Needed: • Product description. Output: • A short list of music/sound keywords (single output). 2) #223 (Powerful LLM Prompt-to-Text Response) Description: Uses segment pacing, brand style, and the brainstormed keywords to: • Suggest one transition type between each labeled segment. • Provide 2–3 overall music genres or moods. • Optionally note 1–2 simple sound effects. Input Needed: • Segments from Subagent 1 and prompts from Subagent 2. • Sound-related keywords from #218. Output: • Transitions for each segment boundary. • 2–3 broad genres/moods. • Optional SFX suggestions (combined into one final output). SUBAGENT REQUIRED INPUT: [segmented-vsl-script], {product-description} SUBAGENT FINAL OUTPUT: [transitions-music-sfx] SUBAGENT 4: Document Compiler Name: Document Compiler Skills Used: • #223 (Powerful LLM Prompt-to-Text Response) Description: Compiles all data (segments, visual prompts, transitions, and music/sound design) into the final text document, "VideoAdCreativeBrief.txt." Input Needed: • Segmented VSL from Subagent 1. • Visual prompts from Subagent 2. • Transition/music from Subagent 3. • Required final format instructions. Output: • "VideoAdCreativeBrief.txt" with labeled segments, 2–3 visual prompts each, transitions, and overall music/sound design. SUBAGENT REQUIRED INPUT: [segmented-vsl-script], [visual-prompts], [transitions-music-sfx] SUBAGENT FINAL OUTPUT: [video-ad-creative-brief]

subagentnumber

[subagentnumber]

Then, agent response...

subagent1refined-new-json-core

I'll add the G) TASKFLOW SUMMARY section to SUBAGENT 1's specification, based on analyzing the tasks/skills described in sections E) and F). -------------- SUBAGENT 1: VSL SCRIPT SEGMENTER A) SUBAGENT SUMMARY This subagent (VSL Script Segmenter) reads the entire VSL script and uses a powerful text-understanding LLM to break it into up to around 6 labeled segments, returning the exact text for each segment. B) FINAL TASK OUTPUT [segmented-vsl-script] → A structured text output in which each segment is labeled (e.g., "Segment 1," "Segment 2," …) and contains the exact script text. C) SUBAGENT INPUT {vsl-script} - The full text of the Video Sales Letter to be segmented. E) SUBAGENT TASK SUMMARY 1) Input: Receive {vsl-script}. 2) Action Step (Skill #223 - Powerful LLM Prompt-to-Text Response): • Prompt the LLM with instructions to identify main messaging or emotional beats in the provided script. • Instruct it to return up to ~6 segments (though fewer or more are acceptable if the script's natural flow warrants it). • Each segment must be labeled clearly ("Segment 1," "Segment 2," etc.). • Each segment must include the exact snippet of the VSL script that it represents. 3) Output: The skill's text response is the final subagent output, named [segmented-vsl-script]. F) SILOS Because this subagent only requires a single LLM action to transform the {vsl-script} into labeled segments, there are no separate silos. The entire segmentation process is contained in a single skill invocation. G) TASKFLOW SUMMARY - ID: #223 - Title task: VSL Script Segmentation and Analysis - Description of the task: Analyze the provided VSL script to identify distinct messaging segments or emotional beats, breaking the script into up to 6 clearly labeled sections while preserving the exact text of each segment. - Input(s) involved: {vsl-script} - Output token name: [segmented-vsl-script] Note: This subagent has a simple, linear workflow with just one primary task/skill (#223) that handles the entire segmentation process. The LLM receives the full VSL script as input and produces a structured output with labeled segments containing the exact script text.

subagent1mermaid

<!DOCTYPE html> <html> <head> <title>Mermaid Diagram Example</title> <script src="https://cdnjs.cloudflare.com/ajax/libs/mermaid/9.4.3/mermaid.min.js"></script> </head> <body> <div class="mermaid"> graph TD A[Start] --> B[Receive VSL Script Input] B --> C[Skill #223: LLM Analysis] C --> D[Process Script Text] D --> E[Identify Key Segments] E --> F[Create Labeled Segments] F --> G[Format Output Structure] G --> H[Generate segmented-vsl-script] H --> I[End] subgraph LLM_Processing C D E F end subgraph Output_Generation G H end </div> <script> mermaid.initialize({ startOnLoad: true }); </script> </body> </html>

Mermaid Image - https://static.aiz.ac/1747486861-mermaid/mermaid1.png

prompts-required

[prompts-required]

subagent1-json-array

[ { "taskID": 223, "semanticTitleOfTask": "VSL Script Segmentation and Analysis", "taskDescription": "This task uses an LLM to break the supplied VSL script into labeled segments (up to ~6) based on natural breaks or emotional beats, ensuring each segment retains the original text exactly.", "inputDescription": "Requires the complete VSL script to analyze and determine ideal segmentation points based on messaging and flow.", "inputRequired": ["vsl-script"], "outputDescription": "Produces a structured text output where each segment is labeled (Segment 1, Segment 2, etc.) and contains the exact corresponding script text.", "outputName": "segmented-vsl-script", "promptInstruction": "You are a Video Ad Creative Director. Analyze the provided VSL script, dividing it into up to ~6 labeled segments. Each segment must contain the exact, unaltered script text. Identify natural breaks based on pacing, emotional beats, or key storytelling points. Label segments as 'Segment 1,' etc. Make sure no text is lost. Format the output:\n\nSegment 1:\nExact script text\n\nSegment 2:\nExact script text\n\nEnsure the segmentation is coherent and covers the entire script." } ]

subagent1-json-array-mermaid

<!DOCTYPE html> <html> <head> <title>Mermaid Diagram Example</title> <script src="https://cdnjs.cloudflare.com/ajax/libs/mermaid/9.4.3/mermaid.min.js"></script> </head> <body> <div class="mermaid"> flowchart TD A[Input: VSL Script Text] --> B[Task #223: VSL Script Segmentation] B --> C{LLM Processing} C --> D[Analyze Script Content] D --> E[Identify Natural Breaks] E --> F[Create Labeled Segments] F --> G[Format Output Structure] G --> H[Output: segmented-vsl-script] H --> I[Segment 1 with exact text] H --> J[Segment 2 with exact text] H --> K[Segment n with exact text] style B fill:#f9f,stroke:#333 style C fill:#bbf,stroke:#333 style H fill:#bfb,stroke:#333 </div> <script> mermaid.initialize({ startOnLoad: true }); </script> </body> </html>